Bio

Hi, my name is Yinhong Liu (刘胤宏). I’m working in Tencent-HY, as a QingYun researcher (青云计划). Previously, I was in Huawei, as a TopMind (天才少年). I had my PhD in Language Technology Lab at the University of Cambridge, working under the supervision of Prof. Nigel Collier.

I’m a fan of Great Automation — the idea that increasingly capable systems can learn, evolve, and improve with minimal human intervention. My interests span the entire lifecycle of self-evolving agents, from data acquisition and autonomous interaction with Human/Environment, to feedback-driven optimization, and continual adaptation. One of my current focuses is Interaction, Full-Duplex, Omni, Agentic interaction model, as I believe it leads to next form of scalable signal.

If you are interested in discussing research or potential collaborations, feel free to reach out via email.

「人生到處知何似，應似飛鴻踏雪泥。」

News!

• 28/09/2026: One paper got accepted by NeurIPS 2026. See you in Sydney!
• 26/08/2025: Three paper got accepted by EMNLP 2025. See you in Suzhou!
• 16/05/2025: Two paper got accepted by ACL 2025, main conference.
• 01/05/2025: One paper got accepted by ICML 2025 as a Spotlight paper. See you in Vancouver.
• 24/01/2025: One paper got accepted by NAACL 2025.
• 26/09/2024: One paper got accepted by NIPS 2024. Great thanks to all the collaborators.
• 21/09/2024: One paper got accepted by EMNLP 2024.
• 10/07/2024: One paper got accepted by COLM 2024. See you in Philadelphia.
• 03/06/2024: I will be interning at Microsoft Research Cambridge.
• 16/05/2024: One paper got accepted by ACL 2024. See you in Bangkok!
• 13/03/2024: We deployed a research paper searching website deep-scholar. Havea try!
• 13/03/2024: One paper got accepted by NAACL 2024, main conference. See you in Maxico City!
• 09/12/2023: I received the [Best Reviewer Award] from EMNLP 2023 generation track.
• 01/06/2023: I am interning at Apple AI/ML this summer.
• 03/11/2022: Going to present the paper Learning Functional Distributional Semantics with Visual Data on Cambridge Language Sciences Annual Symposium 2022.
• 25/10/2022: One paper got accepted by GEM workshop in EMNLP 2022.
• 21/03/2022: One paper got accepted by ACL 2022, main conference. See you in Dublin.

Featured publications

Recursive Self-Improvement & Agents:

• Data-Centric Recursive Improvement for Foundation Models: A Survey Co-author Under Review.

• Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents Zilin Wang, Yinhong Liu, Chao Chen, Qun Liu, Zhijiang Guo. Under Review.

• From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning
  Chao Chen, Chengzu Li, Zhiwei Li, Yinhong Liu, Zhijiang Guo Under Review.

LLM Evaluation & LLM-as-a-Judge:

• Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators
  Yinhong Liu*, Han Zhou*, Zhijiang Guo, Ehsan Shareghi, Ivan Vuli´c, Anna Korhonen and Nigel Collier. [code]
  COLM, 2024.

• MDSEval: A Meta-Evaluation Benchmark for Multimodal Dialogue Summarization.Images link
  Yinhong Liu, Jianfeng He, et al, Hang Su and Saab Mansour
  EMNLP 2025.

• Unlocking Structure Measuring: Introducing PDD, an Automatic Metric for Positional Discourse Coherence
  Yinhong Liu, Yixuan Su, Ehsan Shareghi and Nigel Collier. [code]
  NAACL, main conference, 2024.

Preference Alignment & Reasoning:

• Aligning with Logic: Measuring, Evaluating and Improving Logical Preference Consistency in Large Language Models [Website]
  Yinhong Liu, Zhijiang Guo, Tianya Liang, Ehsan Shareghi, Ivan Vuli´c and Nigel Collier.
  ICML, Spotlight, 2025.

• Operadic consistency: a label-free signal for compositional reasoning failures in LLMs
  Nathaniel Bottman, Yinhong Liu, Kyle Richardson NeurIPS 2026w.

• Improving Preference Extraction In LLMs By Identifying Latent Knowledge Through Classifying Probes
  Sharan Maiya, Yinhong Liu, Ramit Debnath and Anna Korhonen
  ACL 2025, main.

• When Personalization Meets Reality: A Multi-Faceted Analysis of Personalized Preference Learning
  Yijiang River Dong, Tiancheng Hu, Yinhong Liu, Ahmet Üstün and Nigel Collier
  EMNLP 2025.

• Fairer Preferences Elicit Improved Human-Aligned Large Language Model Judgments
  Han Zhou, Xingchen Wan, Yinhong Liu, Nigel Collier, Ivan Vulić, Anna Korhonen [code]
  EMNLP, main, 2024.

Agentic Environment & Data Synthesis:

• Action-Space Shaping for LLM Agents: Measuring and Mitigating Tool-Schema Bias [Code]
  Yinhong Liu, Zhili Tan, Zilin Wang, Zhijiang Guo.
  Under Review.

• EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL Co-author
  Under Review.

• Dial HEALTHDIAL for Advice: A Multilingual and Multi-Parallel Spoken Dialogue Dataset for Knowledge-Grounded Information Seeking
  Songbo Hu*, Yinhong Liu*, Yijie Zhou*, et al, Ivan Vulić** and Anna Korhonen**.
  ACL 2026.

• TOAD: Task-Oriented Automatic Dialogs with Diverse Response Styles [Apple website] [Poster]
  Yinhong Liu, Yimai Fang, David Vandyke and Nigel Collier. [code]
  ACL, findings, 2024.

Others:

• From Chains to Graphs: Self-Structured Reasoning for General-Domain LLMs
  Yingjian Chen, Haoran Liu, Yinhong Liu, et al., Irene Li AACL-IJCNLP 2026.

• Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning
  Jiayu Yang, Chao Chen, Shengen Wu, Yinhong Liu, Yuxuan Fan, Lujundong Li, Songning Lai, Chengwei Qin, Zhijiang Guo EMNLP 2026.

• GraphCheck: Breaking Long-Term Text Barriers with Extracted Knowledge Graph-Powered Fact-Checking
  Yingjian Chen, Haoran Liu, Yinhong Liu, Rui Yang, Han Yuan, Yanran Fu, Pengyuan Zhou, Qingyu Chen, James Caverlee and Irene Li
  ACL 2025, main.

• FASTFACT: Faster, Stronger Long-Form Factuality Evaluations in LLMs
  Yingjia Wan, Haochen Tan, $\cdots$, Yinhong Liu* and Zhijiang Guo*
  EMNLP 2025.

• MR-BEN: A Comprehensive Meta-Reasoning Benchmark for Large Language Models
  Zhongshen Zeng, Yinhong Liu, Yingjia Wan, Jingyao Li, et al. [code]
  NIPS, 2024.

• Prompt Compression for Large Language Models: A Survey
  Zongqian Li*, Yinhong Liu*, Yixuan Su and Nigel Collier.
  NAACL, main, 2025.

• Instruct-SCTG: Guiding Sequential Controlled Text Generation through Instructions
  Yinhong Liu, Yixuan Su, Ehsan Shareghi and Nigel Collier. [code]

• Plug-and-play Recipe Generation with Content Planning
  Yinhong Liu, Yixuan Su, Ehsan Shareghi and Nigel Collier. [code]
  EMNLP, Generation, Evaluation & Metrics (GEM) Workshop, 2022.

• Learning Functional Distributional Semantics with Visual Data[slides]
  Yinhong Liu and Guy Emerson. [code]
  ACL, main conference. 2022.

• Multi-channel audio statistical restoration
  Yinhong Liu and Simon Godsill.
  IEEE International Conference on Signal Processing, Communications and Computing (IEEE-ICSPCC), 2020.

Education

• University of Cambridge, 2021-2026
  PhD in Computation, Cognition and Language. Supervisied by Prof. Nigel Collier and Prof. Ehsan Shareghi.
• University of Cambridge, 2020-2021
  Mphil in Advanced Computer Science. Supervisied by Dr. Guy Emerson.
• University of Cambridge, 2014-2018
  B.A and MEng in Information Engineering. Supervisied by Prof. Simon Godsill.

Professional Service

• Program Committee Member/ Reviewer: ICLR (2024-2025), ACL Rolling Review (2022-2024), ACL (2023-2024), NAACL (2023-2024), EMNLP (2022-2024), Generation, Evaluation & Metrics (GEM) workshop (2022-2023), EACL (2023)
• Teaching:
  TA for Li18 Python for Computational Linguistics, 2021-2023.